your femlab trial: an introduction to femlab's multiphysics

| 1

Y o u r F E M L A B T r i a l :

A n I n t r o d u c t i o n t o F E M L A B ’ sM u l t i p h y s i c s M o d e l i n g C a p a b i l i t i e s

2 | A N I N T R O D U C T I O N T O F E M L A B ’ S M U L T I P H Y S I C S M O D E L I N G C A P A B I L I T I E S

P r e f a c e

. This manual gives you an introduction to modeling in FEMLAB by taking you through

the basic steps of the modeling process, from generating the geometry to postprocessing.

It also provides you with a list of resources for more in-depth technical information.

Enjoy your modeling!

COMSOL, Inc.

8 New England Executive Park

Burlington, MA 01803

Phone: 781-273-3322

Fax: 781-273-6603

www.comsol.com

TE C H N I C A L R E S O U R C E S | 3

T e c hn i c a l R e s ou r c e s

Throughout your trial period, you have access to the following sources of information:

manuals, online tutorials and knowledge base, model library, online command reference

and FEMLAB support team. Here is a brief overview of these resources:

Manuals

Trial users receive the full set of manuals built into the software. To access these manuals,

in FEMLAB go to the Help menu and select FEMLAB Help Desk. We provide the

documentation in PDF and HTML format. These manuals include:

1 User’s Guide and Introduction manual: This manual demonstrates the basics of how

to create a model, how to use the graphical user interface and how to specify a

geometry. It also describes application modes available in the core FEMLAB package

and much more.

2 Model Library manual: This manual documents the models of the model library. These

are problems that have been solved using FEMLAB that we thought might help you

understand the range of problems that FEMLAB can model. These models can also be

used for inspiration, as starting points for your own modeling.

3 Reference manual: This manual contains information on more advanced features of

the software, a description of the various GUI windows, menus and toolbars, as well

as a reference library of all FEMLAB commands.

4 Module manuals: The manual for each module describes the application modes of that

module. You will also find extensive descriptions of the models of the model library

that are based on the module in question.

When you purchase the software you will also receive a set of printed manuals.

Online tutorials

On the webpage:

http://www.comsol.com/showroom/tutorials/

you will find animated tutorials that take you step by step through the set-up of

multiphysics models.


Online Knowledge Base

The webpage:

http://www.comsol.com/support/knowledgebase/

is your first stop for troubleshooting assistance. Search this area of our website if you are

experiencing difficulty with an installation, receiving errors at run-time, or seeing

unexpected behavior.

Here, we have also collected a large number of FEMLAB models provided by users. You

can find tips and tricks as well as unusual solutions to tough modeling situations.

Model library

The model library has been described above. It can be accessed either through the New

item of the File menu or through the Examples and Demos entry of the Help menu. Note

the Documentation button on the Model Library tab of the Model Navigator which gives

you access to the online documentation associated with each model.

Online command reference

While FEMLAB possesses a graphical user interface, you can also create a FEMLAB

model by typing code at the MATLAB command line. These commands are documented

in the Reference manual as explained earlier but you can also get information about a

command from the MATLAB command line. For instance, to learn about the

“postinterp” command, you can type

doc postinterp

or

help postinterp

at the MATLAB command line.

Support team

At any point during your trial, you should feel free to contact our technical staff with your

questions. Do not get stuck with a modeling problem! E-mail the support team at

TE C H N I C A L R E S O U R C E S | 5

[email protected] for help! Alternatively, you can call support at 781-273-3322.

Your trial is also a great opportunity for you to discover the quality of the support you

will receive when you become a customer.


I n t r o du c t i o n

FEMLAB offers a complete modeling environment that allows you to perform all the

steps in the modeling process. FEMLAB’s graphical user interface includes functions for

CAD modeling, import of drawings and images, physics or equation definition, mesh

generation, equation solving, visualization and postprocessing. The parametric solver

gives you a perfect way of examining a parameterized series of models. The parameter

varied typically represents a material property, frequency or reaction rate.

The analysis in engineering design does not end with the computation of voltages,

stresses, and velocity fields. Frequently, we have to perform extensive post processing on

the computed output in order to arrive at useful engineering parameters like impedance,

quality factor, conversion factors, rotation angles etc. In the examples in this primer, you

will be able to translate the simulation results into engineering design numbers.

The following example model gives an overview of basic modeling and introduces the

concepts of multiphysics in the two-way coupling of heat transport and current

conduction in an electronic device. The concepts demonstrated in this example apply to

all types of FEMLAB models, be they from chemical engineering, structural mechanics

or electromagnetics. To demonstrate the various application modes available in

FEMLAB, the heat transfer part of this multiphysics problem is defined first using the

“heat transfer” physics mode (a parametric sweep is demonstrated), then the problem is

solved using the PDE modes and finally we employ the weak mode feature of FEMLAB.

For the PDE modes, both coefficient form and general form are illustrated.

As a last step, we perform a simple parametric study from the Graphical User Interface

and demonstrate the use of the scripting language.

YO U R F I R S T M U L T I P H Y S I C S M O D E L | 7

Y ou r F i r s t Mu l t i p h y s i c s Mode l

Let us start by defining and solving our first model in FEMLAB. The model treats current

conduction and heat generation in an aluminum film deposited on a silicon substrate. The

purpose of the model is to investigate the temperature in the device after the current is

applied. This model is based on a submission by M.S. Cohen at Aegis Semiconductors.

The model exemplifies the important steps in a typical modeling process. It also gives a

quick overrview of some of the features available for advanced modeling in FEMLAB’s

graphical user interface. We will address the following important FEMLAB features:

• The use of predefined physics.

• The definition of a multiphysics problem.

• The expression feature to define physical properties that depend on the solution itself.

• The extraction of design numbers using the postprocessing tools.

• The parameter solver feature that provides fast and efficient parameter screening as

well as smooth convergence for highly nonlinear problems.

In addition to the FEMLAB features, several modeling methods are exemplified in the

exercise, among them:

• Reduction of the dimension of parts of the problem from three to two dimensions.

• Control of the obtained results to estimate the validity of the solutions.

Apart from these highlights, we will go through the typical modeling steps, which

include:

1 Definition of the geometry

2 Definition of the physics in the volume and at the boundaries

3 Meshing

4 Solving

5 Postprocessing

6 Parametric studies

Although in a typical modeling session the analyst would usually try to use the predefined

physics modes as much as possible, we will also illustrate with this model how the PDE

modes and weak modes of FEMLAB can be used to solve the heat transfer problem.


These applications modes are extremely useful when it comes to analyzing non-standard

problems for which no physics modes are available.

General Description of the Physical Problem and Study

A film of aluminum is deposited on a silicon substrate. The aluminum is about 1 µm

thick, with the pattern shown above. The width of the pattern is about 1000 µm, with a

hole in the center about 250 µm in diameter. A constant current is passed through the

deposit starting at time t = 0.

We wish to study the temperature distribution in the aluminum pattern as a function of

time. We also want to see the electrical power distribution as well as the total resistance

as a function of time.

The problem couples two physical phenomena, a current balance and a heat balance. The

electrical resistance increases with temperature while the current density itself serves as

a heat source through ohmic losses in the current conduction. This implies that the

multiphysics coupling between the two physics is a two-way coupling (i.e. we can not

solve either problem separately from the other).

To simplify this problem and keep run-times low, we will look at the current and heat

balances in a two-dimensional approximation of the real geometry. This will give us an


estimation of the time scale of the problem and of the magnitude of the studied fields,

potential and temperature.

A more advanced analysis would consist in solving a full-fledged 3D model of the film.

Such a study could be performed in FEMLAB. Since the purpose of this document is only

to familiarize ourselves with the software, we will not go through this analysis.

C U R R E N T B A L A N C E A N D H E A T B A L A N C E S I N A T W O - D I M E N S I O N A L

C R O S S S E C T I O N

Let us look at the current distribution in the device. Current is only conducted in the

aluminum foil since the silicon substrate is an electrical insulator. Assuming isothermal

conditions, the current density along the thickness should be perfectly uniform, since no

variations are introduced by the geometry, current conduction, or boundary conditions.

This means that we can reduce the problem down to two dimensions to study the current

balance. The current passes through the device from right to left, in the figure above, and

all the boundaries except the left and right boundaries are assumed to be insulated. We

will study the current density distribution, and the influence of temperature on the total

ohmic losses in the device.

The heat balance in the device however is clearly a three dimensional phenomenon. That

being said, we can get an estimate of the production of heat in the device by performing

a two-dimensional analysis. We can introduce the convective cooling perpendicular to the

surface as a heat sink, in order to investigate the convective cooling’s influence on the

time scale of the phenomena. This would give a good start as we take the next step to

three-dimensional modeling.


Below follows a detailed description of the modeling procedure in this exercise. The

step-by-step instructions will guide us through the current balance at steady state and the

introduction of the multiphysics content in the time-dependent heat transfer problem.

NOTE: Although in reality the conductivity of the aluminum layer depends on its temperature, we first neglect this dependence and solve the electricity part of the problem with an assumed constant conductivity. This is meant to help us get familiar with basic FEMLAB concepts with a simpler problem. In a second step we will add this dependence and solve the actual physical problem where the electrical problem and the heat transfer problems are coupled and we need to solve for the electric potential and temperature simultaneously.

The Model Navigator• Start FEMLAB by double-clicking on the FEMLAB icon on the desktop or by typing

femlab in the Matlab command window.

• In the Model Navigator, set Dimension to 2D and then select Physics mode (double

click), Conductive Media DC (double-click), Linear Stationary in the list of available

application modes.

Note: Conductive Media DC is one of a number of physics problems for which

FEMLAB possesses a tailored interface. We refer to those as application modes.

You will find a description of the FEMLAB core application modes in the User’s

Guide and Introduction manual in the chapter called The Application Modes.


Application modes that come in the modules are documented in the module

manuals.

• Press OK.

Draw ModeThe second step in the modeling process is to create or import the model geometry.

FEMLAB is able to read and repair DXF and IGES files from other solid modeling

packages. However, the geometry of the aluminum deposit is very easily created in

FEMLAB’s built-in solid modeling tool.

Note: Extensive information on the various ways of creating a geometry is

available from the Geometry Modeling chapter of the User’s Guide and

Introduction manual.

First we can specify the range of the drawing window and specify the grid spacings.

• Select Axes/Grid settings in the Options menu

• Type -5.5e-5 in the X min and 1.05e-3 in the X max edit field

Note: FEMLAB does not use a specific set of units: you can specify your problem

in any consistent set of units. Quantities output by FEMLAB will then be in the

same set of units. Here we use SI units.

• Type -3.5e-4 in the Y min and 3.5e-4 in the Y max edit field


• Click on the Grid tab

• Clear the Auto box

• Type 1e-4 in the X spacing and Y spacing edit fields

• Insert extra grid lines by typing [-2.5 2.5 5 12.5 17.5]*1e-5 in the Extra Y edit

field and press OK.

This gives a proper grid spacing for the drawing of the device. You can now start drawing

the geometry. The geometry can be created using solid primitive objects or line and curve

tools. We will use the line tool to create the main part of the device.

• Press the Draw Line tool button (It is sixth from the top in the draw toolbar on the left

side of the FEMLAB window.

Note If you hold the cursor steady on a tool button, its associated function is

displayed. Also, note that you can see the location of the pointer in the left most

box of the status bar at the bottom left corner of the FEMLAB window).

• Click on the coordinates tabulated below to create the thick part of the device.

• Press the right mouse button to create a solid object denoted CO1, composite object 1.

X COORDINATES Y COORDINATES

2e-4 0

2e-4 0.5e-4

4e-4 1.75e-4

6e-4 1.75e-4

8e-4 0.5e-4

8e-4 0


• Copy CO1 by using the short-cut key combinations Ctrl-C and Ctrl-V or by

selecting Copy followed by Paste in the Edit menu.

• Type 0 in the X-displacement and Y-displacement edit fields and press OK. Note that

the new object, denoted CO2, is behind CO1 and therefore is indistinguishable.

• Press the Rotate button and type 180 in the Rotation and [5e-4 0] in the Center

coordinates edit fields respectively.

You have created CO1 and CO2 and can continue with FEMLAB’s Create Composite

Object tool to unify these objects.

• Select CO1 and CO2 by using the key combination Ctrl-A and press the Create

Composite Object button (fifth button from the bottom on the vertical toolbar). Check

that the Set formula is CO1+CO2. This means that the resulting composite object is the

union of CO1 and CO2.

• Clear the Keep internal borders box and press OK.

Draw the contact strips to the main body of the device:

• Press the Draw Rectangle tool (first button on the vertical toolbar).


• Hold down the left mouse button and drag the mouse between the coordinates

(0,-0.25e-4) and (1e-3,0.25e-4)to create a rectangle.

• Select the rectangle R1 and the composite CO1 and unify them by using the Create

Composite Object tool. The Keep internal borders box should be cleared.

Most of the geometry is now done. You can now create the circle in the middle and drill

the hole in the structure as follows:

• Select the Draw Centered Ellipse tool.

• Using the right mouse button, create a circle by dragging the mouse straight upwards

from (5e-4,0) to (5e-4,1.25e-4).

Note: You can check that the disc has the proper center and radius by

double-clicking on C1.

• Select all the objects by using the key combination Ctrl-A and press the Difference

button (sixth button from the bottom on the vertical toolbar).

The geometry of the aluminum deposit is completed and you can go to the Boundary

mode:

Boundary Mode• Press the Boundary Mode button on the main toolbar (fifth from right)


• Select all boundaries by using the key combination Ctrl-A

• Select Boundary settings in the Boundary menu

• Set Insulation/symmetry conditions at all boundaries

• Select the right hand side boundary, boundary 16, and press the Inward current density

radio button. You can check that you have selected the correct boundary in the Domain

selection list in the Boundary settings dialog box.

Type iin/(thick*side) in the g edit field. iin, thick, and side are constants that

you will define later.

• Select the left hand side boundary, number 1, and set the Antisymmetry/ground

boundary condition.

• Press OK.

Note: The boundaries are rendered differently based on what boundary conditions

apply there.

The next step is to define the constants used in the above expression in a temporary data

base.

• Select Add/edit constants in the Options menu

• Define the constants tabulated below in the corresponding edit fields. This is done by

entering the name and expression in the table below into the appropriate dialog boxes

and then clicking either the Set or the Apply button. sig is a constant that you will use

later on to define the electrical conductivity of the material.

• Press OK

This defines the boundary conditions for the current density balance. Proceed to the

subdomain mode.

NAME OF CONSTANT EXPRESSION

iin 0.5

thick 1e-6

side 50e-6

sig 3.54e7


Subdomain Mode• Press the Subdomain Mode button in the main toolbar

• Double-click on the geometry or open Subdomain settings from the Subdomain menu

• Type sig in the Conductivity edit field

• Type 0 in the Current source edit field

• Press OK

The constant sig corresponds to the conductivity, in S m-1, at 293 K.

Mesh Mode• Press the Mesh Mode button in the main toolbar to create the mesh

Note: Here we are using the default settings for mesh generation, but very

elaborate meshes can also be created from the Mesh Parameters dialog window,

which is documented in the Graphical User Interface chapter of the Reference

manual. Finer meshes usually yield more accurate solutions and it is always a good

idea to compare the solution obtained with different meshes to make sure that a

sufficiently accurate solution has been reached, as will be done later on in this

document.

Solve Mode• Solve the problem by pressing the Solve Problem button (=) in the main toolbar


Post ModeThe graphical user interface automatically selects the Post Mode once the problem is

solved. The default plot shows the voltage in the aluminum deposit.Post mode quick

buttons

You can toggle between a number of visualization options by pressing the quick buttons

in the Post Mode.

• Press the Plot Parameters button in the main toolbar

• Clear the Surface box and check the Contour box

• Select the Contour tab

• Type 31 in the Contour levels edit field. This will render 31 contour lines with constant

potential increment

• Press OK

Note: The iso-potential lines above and below the hole are perfectly vertical, see

the figure below. This implies that anti-symmetry is present along these vertical

lines while symmetry is obtained between the upper and lower part of the geometry.

The presence of symmetry and anti-symmetry would allow for the modeling of only

one fourth of the device.


Note: While the FEMLAB graphical user interface allows advanced

postprocessing, there are cases where you may need to export your solution to an

other program for analysis of the solution. This is easily done in FEMLAB. Visit

our Knowledge Base at the following URL to familiarize yourself with this:

http://www.comsol.com/support/knowledgebase/

Now that we have familiarized ourselves with the FEMLAB interface and its main

concepts, we can start modeling the actual problem of interest by introducing the heat

transfer problem which, as indicated earlier, is coupled to the electric problem. To

demonstrate FEMLAB’s capabilities, the thermal part of the problem will be specified

consecutively using four different application modes: heat transfer physics mode,

coefficient form PDE mode, general form PDE mode and weak mode.

The next step introduces the concept of multiphysics. FEMLAB allows for arbitrary

coupling of several physical phenomena. In this case, you will add the heat balance to the

existing current balance.


Before we do that, let us define some constants and expressions that will be required for

all four application modes.

First, the constants:

• Select Add/edit constants in the Options menu

• Define the constants according to the table below

• Press OK

In addition to constants, you can define expressions that are functions of the dependent

variables in the model, in this case voltage V and temperature T.

Note 1: The voltage and temperature were named V and T respectively

automatically by FEMLAB when we selected the application modes from the Model

Navigator. If these names had not suited our needs, we could have changed them

from the Model Navigator.

NAME OF CONSTANT EXPRESSION

ro1 2.7e3

cp1 900

k1 240

h1 5

T0 293


Note 2: Expressions can also depend on independent variables (such as x and y)

and include calls to Matlab functions, including user-specified functions.

• Select Add/edit expressions in the Options menu

• Type res in the Variable name edit field

• Press Add

• Click the Definition tab

• Define the ohmic resistivity as a function of temperature according to the table below

by typing in the expression in the corresponding edit field. When you enter a scalar

term like the one below in a FEMLAB dialog box, make sure not to use spaces (In

FEMLAB, spaces are used to separate the entries of a vector).

• Press OK

• Now, save your model under the title FilmBasic.mat using the File menu item Save.

We will reuse this file as a basis for future modeling.

• Save the working model under the file name FilmHTPM.mat using the Save as item of

the File menu. The letters ‘HTPM’ are a reminder that we will be using the Heat

Transfer Physics Mode.

Note: There are two formats in which you can save you FEMLAB models. One is

the .mat format which allows you at a later to time to re-open you model where you

left it, with the computed solution immediately available. The second format is the

.m format which is a script format: the .m file contains a log of all the Matlab

prompt commands that are equivalent to the operations performed in the GUI.

When an .m file is open, all these commands are executed again in sequence. It is

possible to add comments to an .m file just like you would in any other Matlab .m

file.

Multiphysics Mode• Select Add/edit modes in the Multiphysics menu

• Select the Heat transfer mode on the left hand side of the Multiphysics dialog box

• Press the Add Selected Application Mode to Current Model (>>) button

• Select Time-dependent as the Solver type

• Press OK

VARIABLE NAME EXPRESSION

res 2.824e-8*(1+3.9e-5*(T-293))


If you click on the Multiphysics item in the main menu bar, you can see that we have two

application modes selected for our problem and that the currently-active one is the Heat

Transfer application mode we just added.

Boundary Mode• Press the Boundary Mode button

• Select all boundaries by typing the key combination Ctrl-A

• Select Boundary settings in the Boundary menu

• Press the radio button to set the heat flux, which is the first button in the Boundary

coefficients list

• Type h1 in the heat transfer coefficient edit field

• Type T0 in the External temperature edit field

• Select the left and right hand side boundaries, boundaries 1 and 16. You can also do

this by holding down the Ctrl key and clicking on the boundaries in the drawing of the

device if you do not know the numbers.

• Press the Temperature radio button and type T0 in the corresponding edit field

• Press OK

You will now continue by defining the domain physics for the heat transfer mode.

Subdomain Mode• Press the Subdomain Mode button

• Double-click on the domain


Note: The equation solved for in the subdomain is displayed at the top of the

Subdomain Settings window.

The heat transfer through convection on the surface of the aluminum deposit is

introduced as a source or a sink. Since invariance along the thickness of the device

implies that the model is defined per unit length in the z direction, the heat transfer

coefficient has to be divided by the thickness of the deposit. This gives the correct value

of the cooling per unit volume of aluminum. The coupling between the heat balance and

the current balance is present in the expression for the resistivity, denoted res, and the

source term, which depends on the gradient of the potential, whose components are

denoted by Vx and Vy. The equation we want to solve for in the subdomain is therefore:

With this in mind, define the Subdomain settings according to the table below:

• Click the Init tab and type T0 in the Initial value edit field

• Press OK

In order to make a two-way coupling between the heat and current balances, you have to

introduce the expression for the resistivity, res, in the conductive media application.

• Select 1Geom1:Conductive media DC in the Multiphysics menu

• Select Subdomain settings in the Subdomain menu

• Type 1/res in the Conductivity edit field instead of sig. The resistivity res was

specified earlier as an expression that depends on the temperature

• Press OK

Solve Mode• Select Parameters in the Solve menu

• Check that the Time dependent radio button in the General page is selected

DESCRIPTION VALUE

Density ro1

Heat capacity cp1

Thermal conductivity k1

Heat source 1/res*(Vx^2+Vy^2)

Convect. heat transf.coeff. h1/thick

External temperature T0

ρ1cp1∂T∂t------ ∇ k1∇ T( )•–

∂V∂x------

2 ∂V∂y------

2+

res------------------------------------

h1

thi ck------------- T0 T–( )+=


• Click the Timestepping tab and type 0:0.001:0.03 in the Output times edit field

• Press OK

The vector expression for the output times gives the solution from 0 to 0.03 seconds with

0.001 second increments. The time stepping algorithm has built-in step length control

and the output time only defines the time points that should be saved for post processing.

• Press the Solve button in the main tool bar

The solution shows the potential distribution in the aluminum deposit for the resulting

temperature distribution. You can plot the conductivity as a function of temperature in the

geometry:

• Press the Plot parameters button in the main toolbar

• Click the General tab

• Clear the Contour box and check the Surface box

• Select the Surface tab


• Type 1/res in the Surface expression edit field

• Press OK

You can study the development of the temperature profile by animating the solution

through the Animate tab of the Plot Parameters window or by plotting the temperature as

a function of time in the hottest part of the device.

• Select Cross-Section plot parameters in the Post menu

• Select the Point tab and press the Point plot radio button

• Select temperature (T) in the Point expression combo box

• Type 500e-6 in the X: and 125e-6 in Y: edit fields

• Select the General tab


• Select all time steps in the Solution at time list

• Press OK

The resulting plot shows that the problem has reached steady-state after 0.02 seconds.

You can continue visualizing the current density, potential, and current density

distribution by combining cross sectional, surface, contours, and flow plots.

C O N V E R G E N C E A N A L Y S I S A N D P O S T- P R O C E S S I N G

To check the numerical accuracy of the solutions, we need to examine the sensitivity of

the values to the mesh density. We will look at a couple of key numbers in the solution at

the end time and compare them at different mesh densities.

• Change to a stationary problem by selecting the Solve/Parameters... menu and under

the General tag select Solver type to Stationary nonlinear and Solution form to General.

• Solve the problem by clicking the = button.

• Evaluate the norm of the temperature gradient in (97e-5, 0) by selecting Post/Cross

Section Plot Parameters from the Post menu and selecting Point on the General tab.

On the Point tab, select Point expression: temp. gradient (gradT_ht) and

Coordinates: X: 97e-5, Y: 0.

• Evaluate the mean temperature

by selecting Subdomain Integration in the Post menu and specifying Expression to

T⟨ ⟩ T A A⁄d∫=


temperature (T). The result will show up in the status field. Divide this number

by the total area, which can be obtained by a second subdomain integration, typing in

1 in the Expression field. The average temperature should be approximately 830.5

degrees.

• The last two steps can be automated with a few lines of MATLAB code. Create an

m-file in C:\MATLAB6p5\FEMLAB23 with the following contents:

grad1=postinterp(fem,'gradT_ht',[97e-5;0])Tmean=postint(fem,'T')/postint(fem,'1')

Note: postinterp and postint are FEMLAB commands that are documented in

the Function Reference chapter of the Reference manual. You can also learn more

about the syntax of these commands by typing

doc postint

or

doc postinterp

at the MATLAB command line.

• In the GUI, do CTRL-F before running the m-file. This will export all of the data

structure from FEMLAB to MATLAB and therefore enable post-processing of the

FEMLAB solution from the MATLAB prompt.

If you wish, you can repeat the runs of the m-file for different mesh densities: select the

Mesh/Parameters... menu and specify the mesh size: More>>, Max element size for

subdomains: 1 1e-4. Repeat to generate the following table.

*Use the following expression in Mesh size expression:

(x<2e-4)*.4e-5+(x>=2e-4)*(x<8e-4)*1.5e-5+(x>=8e-4)*.4e-5

The following MATLAB code is used to generate two convergence diagrams:

Tmeans=[830.4682 830.5551 831.7698 832.4117 832.6016];gradTs=[3.4428e6 3.4430e6 3.4474e6 3.4496e6 3.4503e6];Mesh=[1e-4 3e-5 1.5e-5 0.75e-5 0.4e-5];figure

MESH ELEMENT SIZE GRADT [97E-5, 0] <T>

1 1e-4 3.4428e6 830.4682

1 3e-5 3.4430e6 830.5551

1 1.5e-5 3.4474e6 831.7698

1 0.75e-5 3.4496e6 832.4117

-* 3.4503e6 832.6016


plot(-log10(Mesh),gradTs)xlabel('-log10(h)')ylabel('gradT')figureplot(-log10(Mesh),Tmeans)xlabel('-log10(h)')ylabel('Tmean')

This series of commands draws the the figure below where piecewise linear interpolation

is employed to join the 5 data points presented in the table on the previous page. The fact

that these diagrams flatten out when the elements get smaller is an indication that we have

probably reached a reasonably accurate solution with the finer meshes and that the

solution would not vary significantly if we refined the mesh further.

This concludes our analysis based on the Heat Transfer physics mode. Save your

simulation by selecting Save from the File menu.

Coefficient Form PDE Mode

We will now solve the exact same problem by use of the Coefficient Form PDE mode for

the heat transfer part of the problem. This is done to demonstrate the Coefficient Form

PDE mode. When there exists a physics mode for the problem you want to model, using

that physics mode is by far the easiest and safest way to go as the user interface is then

tailored for the problem in question. However, when you wish to solve an equation for

which there is no physics mode available the PDE modes allow you to specify your own

equation. The Coefficient Form PDE mode is most appropriate for modeling linear or

weakly nonlinear problems.


Note: the PDE modes are documented in the chapter of the User’s Guide and

Introduction manual entitled The FEMLAB PDE Language.

• First, open the file FilmBasic.mat by selecting Open in the File menu

• Save the model as FilmCoeff.mat using the Save as item in the File menu

• Select Add/edit modes in the Multiphysics menu

• Highlight the mode named PDE, coefficient form on the left hand side of the

Multiphysics dialog box and name T the dependent variable.


• Change the Solver type to Time dependent

• Press OK

Note: If you click on the Multiphysics item on the main menu bar, you can check

that two application modes are selected and the active one is the Coefficient Form

PDE mode.

Subdomain Mode• Select Subdomain Settings from the Subdomain menu and click on subdomain 1.

The equation we need to solve in the subdomain is still

The equation solved for in the Coefficient Form PDE mode is

Note: This equation is displayed at the top of the Subdomain Settings window.

• Therefore, enter the data from the following table in the dialog fields of the Subdomain

Settings window after selecting domain 1 from the Domain selection list:

DESCRIPTION VALUE

c k1

a h1/thick

f (1/res*(Vx^2+Vy^2))+(h1/thick)*T0

da ro1*cp1

ρ1cp1∂T∂t------ ∇ k1∇ T( )•–

∂V∂x------

2 ∂V∂y------

2+

res------------------------------------

h1

thi ck------------- T0 T–( )+=

da∂T∂t------ ∇ c∇ T αT γ–+( ) aT β ∇ T•+ +•– f=



• Press OK

Boundary ModeWe want to impose Neumann boundary conditions on all boundaries except 1 and 16 and

Dirichlet conditions on those two. The easy way to do this is to first impose Neumann

boundary conditions on all boundaries and then modify the boundary conditions applied

to 1 and 16 (Same as we did earlier).

• Press the Boundary Mode button (fifth from right on the main tool bar, below the menu

bar)

• Select all boundaries using Ctrl+A

• Select Boundary Settings from the Boundary menu. Make sure to toggle on the

Neumann boundary condition. As you can see at the top of the window, the Neuman

boundary condition in the Coefficient Form PDE mode is:

0 0

0 0

0 0

DESCRIPTION VALUE

αβϒ

n c∇T αT ϒ–+( )• qT+ g=


The boundary condition we need to impose is as before:

We have already defined c, and . Enter the values in the table below into the proper

dialog fields:

• Select boundaries 1 and 16 from the Domain selection list (holding the control key)

• Select Dirichlet boundary condition

The Dirichlet boundary condition in the Coefficient Form PDE mode is:

The rightmost term in the first of these two equations corresponds to a Lagrange

multiplier used to enforce the Dirichlet condition.

• Enter the values in the table below into the proper dialog fields

• Click OK

Subdomain ModeBefore we run the simulation, we need to finish specifying the two-way coupling by

setting the electric conductivity to 1/res in the Subdomain Settings of the electrical

application mode.

• Select I Geom1: conductive media DC (dc) from the Multiphysics menu to switch back

to the electrical application mode

• In the Subdomain Settings (under the Subdomain menu) window, enter 1/res instead

of sig as the conductivity for subdomain 1

• Click OK

DESCRIPTION VALUE

q h1

g h1*T0

DESCRIPTION VALUE

q 0

g

h

r

0

1

T0

n k1∇T( )• h1 T0 T–( )=

α ϒ

n c∇T αT ϒ–+( )• qT+ g hTµ–= ; hT r=



• Check that the Time dependent radio button is selected on the General page


• Press OK


The solution is the same as earlier. For instance, you could easily verify that the same

integral for the temperature is obtained, with a value of 0.00010873.

• Save your model by selecting Save from the File menu. Alternatively, use Ctrl+S

General Form PDE Mode

We will now solve the exact same problem by use of the General Form PDE mode for the

heat transfer part of the problem. Our goal is to demonstrate this alternative way of

specifying your own equations in FEMLAB. This form is particularly appropriate for

non-linear problems.

Note: the PDE modes are documented in the chapter of the User’s Guide and

Introduction manual entitled The FEMLAB PDE Language.


• Save the model as FilmGeneral.mat using the Save as item in the File menu


• Highlight the mode named PDE, general form on the left hand side of the Multiphysics

dialog box and name T the dependent variable

• Change the Solution form to General


• Change the Solver type to Time dependent

• Press OK

• Note that if you click on the Multiphysics item on the main menu bar, you can check

that two application modes are selected and the active one is the General Form PDE

mode


Subdomain Mode• Select Subdomain Settings from the Subdomain menu and click on subdomain 1

Once again, the equation we need to solve in the subdomain is:

The equation solved for in the General Form PDE mode is:

Observe that this equation is displayed at the top of the Subdomain Settings window.

• Therefore, enter the data from the following table in the dialog fields of the Subdomain

Settings window:

It is important that you leave a space between -k1*Tx and -k1*Ty to specify that these

are the two components of the vector.


• Press OK

DESCRIPTION VALUE

-k1*Tx -k1*Ty

F (1/res)*(Vx^2+Vy^2)+(h1/thick)*(T0-T)

da ro1*cp1

ρ1cp1∂T∂t------ ∇ k1∇ T( )•–

∂V∂x------

2 ∂V∂y------

2+

res------------------------------------

h1

thi ck------------- T0 T–( )+=

da∂T∂t------ ∇ Γ•+ F=

Γ

Γ



Dirichlet conditions on those two. As earlier, we will do this by first imposing Neumann

boundary conditions on all boundaries and then modifying the boundary conditions

applied to 1 and 16.

• Press the Boundary Mode button


• Select Boundary Settings from the Boundary menu. Make sure that the Neumann

boundary condition is toggled on. As you can see at the top of the window, the

equation for boundary condition in the Coefficient Form PDE mode is:

The boundary condition we need to impose is as before:

We have already defined .

• Enter h1*(T0-T) in the G dialog field

• Select boundaries 1 and 16 from the Domain selection list

• Select Dirichlet boundary condition

• Enter 0 in the G dialog field and -T+T0 in the R dialog field

• Click OK

DC Subdomain ModeBefore we run the simulation, we need to finish specifying the two-way coupling by

setting again the electric conductivity to 1/res in the Subdomain Settings of the

electrical application mode.





• Click OK


n Γ•– G∂R∂T------

Tµ+= ; R 0=

n k1∇T( )• h1 T0 T–( )=

Γ


• Press the Time dependent radio button in the General page


• Press OK


The solution is the same as earlier. For instance, you could easily verify that the same

integral for the temperature is obtained (Again, 0.00010873 should be its value).

• Save your model by selecting Save from the File menu. Alternatively, use Ctrl+S.

Weak Mode

We will now solve the exact same problem by use of the Weak mode for the heat transfer

part of the problem. The Weak mode is a very general way of specifying your governing

equation in FEMLAB. It offers a great deal of flexibility and may be required in the

creation some very advanced models.


• Save the model as FilmWeak.mat using the Save as item in the File menu


• Highlight the mode named Weak, subdomain on the left hand side of the Multiphysics

dialog box and name T the dependent variable


• Change the Solver type to Time-dependent

• Press OK

Note: that if you click on the Multiphysics item on the main menu bar, you can

check that two application modes are selected and the active one is the Subdomain

weak form PDE mode.

Subdomain Mode• Select Subdomain Settings from the Subdomain menu and click on subdomain 1.

The equation we need to solve in the subdomain is

ρ1cp1∂T∂t------ ∇ k1∇ T( )•–

∂V∂x------

2 ∂V∂y------

2+

res------------------------------------

h1

thi ck------------- T0 T–( )+=


Multiplying by a test function , integrating over the subdomain, integrating by parts the

diffusion term and rearranging terms yields:

where denotes the part of the boundaries where Neumann conditions are imposed

and the test function is chosen to vanish where Dirichlet conditions are satisfied by the

solution.

• Select subdomain 1 from the Domain selection list

• Enter -k1*(Tx_test*Tx+Ty_test*Ty)+T_test*((1/

res)*(Vx^2+Vy^2)+(h1/thick)*(T0-T)) in the weak dialog field

• Enter T_test*ro1*cp1*T_time in the dweak dialog field


• Click OK


Dirichlet conditions on those two. As earlier, we do this by first imposing Neumann

T

Tρ1cp1∂T∂t------

Ω∫ …=

… k1∇T ∇ T T

∂V∂x------

2 ∂V∂y------

2+

res------------------------------------

h1

thi ck------------- T0 T–( )+

Th1 T0 T–( )∂ΩN

∫+ +•–

Ω∫=

∂ΩN


boundary conditions on all boundaries and then modifying the boundary conditions

applied to 1 and 16.

• Press the Boundary Mode button


• Select Boundary Settings from the Boundary menu

• In the weak dialog field, enter T_test*h1*(T0-T)

• Enter 0 in the constr dialog field

• Select boundaries 1 and 16 from the Domain selection list

• Enter 0 in the weak field and -T+T0 in the constr dialog field

• Click OK

DC Subdomain ModeBefore we run the simulation, we need to finish specifying the two-way coupling by again

setting the electric conductivity to 1/res in the Subdomain Settings of the electrical

application mode.





• Click OK


• Press the Time dependent radio button in the General page


• Press OK


The solution is the same as earlier. Once again, you can check that the same integral as

before is obtained for the temperature.

Save your model by selecting Save from the File menu. Alternatively, use Ctrl+S.


Parametric Study

We conclude our study of this problem with a simple parametric study. In the original

electric problem uncoupled to the heat transfer problem, we considered a uniform and

constant conductivity sig. Now, we will consider a conductivity that is not uniform but

instead varies in the form sig+SigVar*x^2. We will look at several values from 0 to

1e13 for the parameter SigVar.

• Open the file FilmBasic.mat from the Open item in the File menu

• Save as FilmParam.mat (under the File menu)

• On the Subdomain Settings page under the Subdomain menu, specify

sig+SigVar*x^2 for the conductivity in subdomain 1

• Go to the General tab on the Solve, Parameters window

• Select Parametric as the Solver type

• Under the Parametric tab, enter SigVar as the Name of parameter and

[0:1e12:1e13] for the List of parameter values. This specifies that we will solve the

problem for 11 equally-spaced values of SigVar between 0 and 1e13.

• Press Solve

• In Post mode, you can plot the solution at any of the specified values of the parameter

SigVar. Simply select that value from the Parameter value list under the General tab

on the Plot parameters window.


Note: An other easy way to visually compare the solutions obtained with different

values of the parameter consists in using the Animation feature in the Plot

parameters window.

Let us plot the voltage at the point of coordinates (5e-4; 1.25e-4). To do so, in the Post

menu, select Cross Section Plot Parameters and under the General tab select all timesteps

(using Ctrl+A). Under the Point tab, toggle Point plot, select V as the point expression

and finally enter 5e-4 as the X coordinate of the point and 1.25e-4 as the Y coordinate.

Click OK.

Note on command line scripting

Whereas we have now created and run all the models above from the FEMLAB Graphical

User Interface, it should be noted that in fact the same studies could have been performed

entirely from the MATLAB command line, through scripting.

There are a number of reasons one might want to use scripting. For one, a script allows

you to readily create several problems that are similar to each other in a minimal amount

of time. Repetitive tasks can be automatized in a script. Also a script can include any

Matlab command, including tests, loops, etc, which means you can base the definition of

a problem on the output on another, automatically. Finally, certain functionalities that are

not implemented in the GUI can be accessed only from the command line, or you may

even add functionalities specific to your application.


For an example of such a script, in the GUI, select Reset model m-file from the File menu.

This will generate a script that generates the input for the current model. Note that this

m-file does not contain the mesh and solution.

For more information on the syntax of FEMLAB commands, it is always possible to type

help commandname or doc commandname at the MATLAB prompt. Extensive

information is also available from the FEMLAB Manuals, which you can access from the

FEMLAB Helpdesk in the Help menu.

As a sample problem specified from the command line, type the following at the

MATLAB prompt:

clear fem;fem.geom=rect2;fem.equ.c=1;fem.equ.f=1;fem.bnd.h=1;fem.bnd.r=0;fem.mesh=meshinit(fem);fem.xmesh=meshextend(fem);fem.sol=femlin(fem);postsurf(fem,’u’)

These lines of code illustrate the solution of the equation on the unit square

with the boundary condition on the boundary.

∇ ∇ u• 1=

u 1=

your femlab trial: an introduction to femlab's multiphysics

Documents